Elsevier

Annals of Epidemiology

Volume 24, Issue 8, August 2014, Pages 575-580.e1
Annals of Epidemiology

Original article
Mortality inequality in populations with equal life expectancy: Arriaga's decomposition method in SAS, Stata, and Excel

https://doi.org/10.1016/j.annepidem.2014.05.006Get rights and content

Abstract

Purpose

Life expectancy is used to measure population health, but large differences in mortality can be masked even when there is no life expectancy gap. We demonstrate how Arriaga's decomposition method can be used to assess inequality in mortality between populations with near equal life expectancy.

Methods

We calculated life expectancy at birth for Quebec and the rest of Canada from 2005 to 2009 using life tables and partitioned the gap between both populations into age and cause-specific components using Arriaga's method.

Results

The life expectancy gap between Quebec and Canada was negligible (<0.1 years). Decomposition of the gap showed that higher lung cancer mortality in Quebec was offset by cardiovascular mortality in the rest of Canada, resulting in identical life expectancy in both groups. Lung cancer in Quebec had a greater impact at early ages, whereas cardiovascular mortality in Canada had a greater impact at older ages.

Conclusions

Despite the absence of a gap, we demonstrate using decomposition analyses how lung cancer at early ages lowered life expectancy in Quebec, whereas cardiovascular causes at older ages lowered life expectancy in Canada. We provide SAS/Stata code and an Excel spreadsheeet to facilitate application of Arriaga's method to other settings.

Introduction

Life expectancy is a useful measure of population health for its capacity to summarize mortality in a single measure. Life expectancy has unique mathematical properties that often go unrecognized which can be tapped to facilitate population comparisons by epidemiologists. The difference in life expectancy between two groups is an algebraic function of underlying age and cause-specific mortality rates [1], [2], [3], [4], [5], [6]. Recent methods proposed in demography have taken advantage of these relationships to show that life expectancy gaps can be partitioned into age and cause-specific components. Such methods have successfully identified the age groups and causes of death resulting in socioeconomic [7], [8], [9], [10], [11], ethnocultural [8], [12], [13], [14], [15], and temporal [7], [8], [10], [11], [13], [14], [15], [16] inequalities in life expectancy in several countries.

Although there is an abundance of decomposition analyses in the literature, such studies are rarely performed in the absence of a life expectancy gap. Populations can have large inequalities in age or cause-specific mortality, yet very similar life expectancy—in such settings, decompositions can still be undertaken to determine the reason for the absence of a gap. No matter how small, the difference in life expectancy between two populations is the sum of positive and negative contributions of age and cause-specific mortality rates [2], [3], [4], [5]. Contributions can be quite large, yet be in opposite directions that cancel each other when summed, potentially yielding little or no life expectancy gap (i.e., equal life expectancy in both groups). For example, life expectancy in a population with mortality that is high at younger and low at older ages could be very similar to another with mortality low at younger and high at older ages, despite the large inequality in age-specific mortality. Decomposition methods are ideal for determining whether life expectancy gaps that appear negligible are in fact associated with large differences in mortality. If the ages or causes of death are sensitive to preventive interventions, such as tobacco cessation, decomposition approaches can provide public policy initiatives with evidence on which age groups or causes of death to target for reducing mortality.

Our primary objective was to demonstrate the utility of decomposition methods for unmasking potentially hidden mortality inequalities among populations with similar life expectancies. Doing so could provide evidence on whether specific age groups or causes of death could be targeted to increase life expectancy. We analyzed mortality data from Canada. Life expectancy of Canadians is high at 79 years for men and 83 years for women, and gaps between provinces are negligible [17]. Yet the province of Quebec has an exceptionally high prevalence of tobacco consumption [17], and higher smoking-related mortality should intuitively lead to lower life expectancy in this province. Decomposition analysis can determine whether tobacco-related mortality indeed affects life expectancy between Quebec and the rest of Canada. Our secondary objective was to encourage similar analyses by researchers and policy makers in other milieus through readily adaptable syntax and an easy-to-use spreadsheet containing formulae for the decomposition of a life expectancy gap.

Section snippets

Data and variables

We obtained data on all deaths in men and women in Quebec and the rest of Canada (hereafter, Canada) for 2005 through 2009 from the Quebec Health and Social Services Ministry and Statistics Canada. Death counts were available in 5-year age blocks (<1, 1–4, 5–9, …, 85–89, and ≥90 years). Population counts were obtained from the Census of Canada (administered in 2006, with estimates for 2005 and 2007–2009) for ages 1 year or more, and birth registration certificates for infants younger than 1

Results

There was essentially no gap in life expectancy between Quebec and Canada (Table 1). The difference was only 0.1 years, for both men and women. There were differences in age and tobacco-related mortality between Quebec and Canada, but the reason why life expectancy was similar between the two areas was not readily apparent from the rates alone (Table 2).

There was no clear pattern in age-specific contributions to the life expectancy gap between Quebec and Canada for women, although higher

Discussion

In this study, we decomposed the life expectancy gap between Quebec and Canada into age and tobacco-related causes of deaths. We selected these two populations for their similar life expectancy, despite differences in prevalence of major risk factors for mortality including tobacco [17]. We aimed to demonstrate that, no matter how small, life expectancy gaps can be partitioned to determine whether risk factors such as tobacco indeed influence life expectancy. Using Quebec and Canada as an

Acknowledgment

This work was supported by a grant from the Research Center of the University of Montreal Hospital Center. N.A. and S.H. acknowledge career awards from the Fonds de recherche du Québec. The authors thank Karine Garneau for programming the Arriaga formulae in Excel.

References (33)

  • S. Yang et al.

    The changing gender differences in life expectancy in Korea 1970-2005

    Soc Sci Med

    (2012)
  • E. Breton et al.

    Health promotion research and practice require sound policy analysis models: the case of Quebec's Tobacco Act

    Soc Sci Med

    (2008)
  • E. Andreev et al.

    Algorithm for decomposition of differences between aggregate demographic measures and its application to life expectancies, healthy life expectancies, parity progression ratios and total fertility rates

    Demogr Res

    (2002)
  • E.E. Arriaga

    Measuring and explaining the change in life expectancies

    Demography

    (1984)
  • E.E. Arriaga

    Changing trends in mortality declines during the last decades

  • J.H. Pollard

    The expectation of life and its relationship to mortality

    J Inst Actuaries

    (1982)
  • J.H. Pollard

    On the decomposition of changes in expectation of life and differentials in life expectancy

    Demography

    (1988)
  • Sex differentials in life expectancy and mortality in developed countries: an analysis by age groups and causes of death from recent and historical data

    Popul Bull UN

    (1988)
  • J. Adam-Smith et al.

    Causes of widening life expectancy inequalities in Quebec, Canada, 1989-2004

    Can J Public Health

    (2011)
  • N. Auger et al.

    Diverging socioeconomic inequality in life expectancy of Francophones and Anglophones in Montréal, Québec: tobacco to blame?

    J Public Health

    (2013)
  • Y.H. Khang et al.

    Decomposition of socio-economic differences in life expectancy at birth by age and cause of death among 4 million South Korean public servants and their dependents

    Int J Epidemiol

    (2010)
  • P. Martikainen et al.

    Change in male and female life expectancy by social class: decomposition by age and cause of death in Finland 1971-95

    J Epidemiol Community Health

    (2001)
  • L. Tarkiainen et al.

    Trends in life expectancy by income from 1988 to 2007: decomposition by age and cause of death

    J Epidemiol Community Health

    (2012)
  • N. Auger et al.

    Life expectancy gap between the Francophone majority and Anglophone minority of a Canadian population

    Eur J Epidemiol

    (2012)
  • S. Harper et al.

    Trends in the black-white life expectancy gap, 2003-2008

    JAMA

    (2012)
  • W. Na'amnih et al.

    Trends in the gap in life expectancy between Arabs and Jews in Israel between 1975 and 2004

    Int J Epidemiol

    (2010)
  • Cited by (50)

    • Inequalities in life expectancy by educational level and its decomposition in Barcelona, 2004-2018

      2022, Gaceta Sanitaria
      Citation Excerpt :

      The total contribution by causes of death is simply obtained by adding the contributions across age groups. We adapted the life tables and decomposition programmes from Auger et al.15 using Stata version 15 (code in the Supplementary Material). A linear regression was performed to identify whether there was a linear tendency in life expectancy and to determine the contribution of age and cause of death in the period analysed.

    • The contribution of alcohol-related deaths to the life-expectancy gap between people with and without depression – a cross-country comparison

      2022, Drug and Alcohol Dependence
      Citation Excerpt :

      To assess the contribution of alcohol-related deaths, the life-expectancy gap between those with and without depression was decomposed by cause of death using the Arriaga decomposition method (1984). Abridged period life tables an decomposition were calculated using Stata version 16 (StataCorp, 2019), adapting the decomposition syntax provided by Auger et al. (2014). The 95% confidence intervals were calculated by bootstrapping.

    View all citing articles on Scopus
    View full text